The Unassuming Economist

From a paper by Antoine Gaudin, Brendan Harnoys-Vannier, and Martin Kessler:

“In the context of the ongoing review of the Debt Sustainability Analysis (DSA) for Low-Income
Countries (LICs), this paper seeks to help shed light on IMF and World Bank macroeconomic
projections. DSAs are central to the financial architecture of developing countries. Yet, the ways the
projections are performed are rarely accessible to outside researchers.

The first contribution of this paper is to provide a newly constructed database of 605 DSAs
conducted from 2013 to 2024. It contains all the information of all published DSAs for LICs in Tables 1
(macro-economic and fiscal) and 2 (external debt dynamics), as well as the shock scenarios. It will be
updated regularly.

The second contribution of the paper is to analyze forecast errors concerning public and external
debt, as well as the main macroeconomic components. It highlights results on large optimistic biases,
with a 10 percentage point underestimation of the trajectory of the debt-to-GDP ratio on average after
5 years. Decomposing this result, it finds that:

1. Larger countries tend to be more affected by significant positive biases. Small Island
   Developing States (SIDS) and vulnerable countries tend to be more accurately forecasted. We
   interpret this finding as showing the integration of past shocks in the baseline.
   
2. It’s mostly fiscal: The main driver of forecast errors is the underestimation of primary deficits, followed by overestimated GDP growth. In particular, forecasting errors on primary deficits stem from overestimated fiscal revenues.
   
3. Mixed results post-2017 reform: While the 2017 reform introduced tools aimed at enhancing forecast realism, biases have persisted. This is evidence of some limited (non-statistically significant) improvements by reducing the optimism bias. This pleads for further disclosure of assumptions. However, given that they were rolled out in 2018, and that COVID-19 made projections difficult, we also caution against too broad interpretation of those results.
   
4. DSAs designed in the context of programs perform better on public debt, but worse on deficits: This tends to show that the IMF tends to overestimate the political feasibility of a program. We find some support for the idea that in LICs, the multipliers are still underestimated.
   
5. Influence of country-specific factors: The study identifies institutional, structural, and cyclical factors influencing these biases, including governance quality, economic diversification, and global economic conditions. Countries reliant on commodity exports tend to have significant forecast biases, particularly optimistic projections for both public and external debt ratios. Countries with fragile governance or in conflict display more pessimistic forecasts for primary deficits and external debt, but overly optimistic growth projections. Countries that have had market access and have build-up debt stocks toward defending market access.
   
6. Recession conditions: DSAs conducted during recessions are associated with strong optimism in public and external debt ratios as well as real GDP growth. This suggests both a tendency to overplay rebound effects and a misconception of the way macroeconomic effects transmit over various phases of the business cycle.
   
   
   Optimism bias is very hard to control, but it can have large policy consequences on the IMF and its members. By publishing more information on its DSAs, the IMF and the World Bank have allowed outside scrutiny. The database we are publishing hopefully provides the tools to outside researchers to help this scrutiny, and we hope that this paper is a first example of such exploration.